i''m seeing issues with how caching interacts with auto const loading
in production. it''s find if the objects cached are top-level
themselves, but an object graph (model + associations) is cached i''ll
get errors when a production process first comes up (no models loaded)
and hits the cache to get said object graph.
this fix, which is a uber hack, shows quite clearly the bug - i''m
*really* curious if anyone else has seen this behaviour
the hack:
vendor/rails/activesupport/lib/active_support/cache/mem_cache_store.rb
24 def read(key, options = nil)
25 super
26 @data.get(key, raw?(options))
27 rescue MemCache::MemCacheError => e
28 logger.error("MemCacheError (#{e}): #{e.message}")
29 nil
30 rescue ArgumentError => e
31 message = e.message
32 raise unless message =~ %r|undefined class/module|
33 class_name = message.scan(%r/\w+/).last
34 begin
35 ActiveSupport::Dependencies.load_missing_constant
ActiveRecord::Base, class_name
36 rescue Object
37 raise e
38 end
39 end
in case it isn''t obvious this simply shows that rails const loading
mechanism is not triggered from a Marshal.load, which makese sense
actually....
thoughts?
a @ http://codeforpeople.com/
--
we can deny everything, except that we have the possibility of being
better. simply reflect on that.
h.h. the 14th dalai lama
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"Ruby on Rails: Core" group.
To post to this group, send email to rubyonrails-core@googlegroups.com
To unsubscribe from this group, send email to
rubyonrails-core+unsubscribe@googlegroups.com
For more options, visit this group at
http://groups.google.com/group/rubyonrails-core?hl=en
-~----------~----~----~----~------~----~------~--~---
> this fix, which is a uber hack, shows quite clearly the bug - i''m > *really* curious if anyone else has seen this behaviourThis has been reported a few times before, but given that it only happens during the initial requests, it''s less likely to be noticed.> in case it isn''t obvious this simply shows that rails const loading > mechanism is not triggered from a Marshal.load, which makese sense > actually....Yeah, marshal doesn''t trigger any of the const_missing hooks that we rely on for dependency loading. There''s not a lot we can do here.> thoughts?If you''re using 2.2 or later the preloading of classes and templates will catch 99.999% of this now as all files in lib/ app/models and app/controllers are required before the first request is processed. -- Cheers Koz --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Ruby on Rails: Core" group. To post to this group, send email to rubyonrails-core@googlegroups.com To unsubscribe from this group, send email to rubyonrails-core+unsubscribe@googlegroups.com For more options, visit this group at http://groups.google.com/group/rubyonrails-core?hl=en -~----------~----~----~----~------~----~------~--~---
cache_fu has a method for handling this type of issue which has been adapted for Rails.cache here: http://github.com/actsasflinn/snippy/blob/77bdc5760180a5b5a8652af6dc6b1e3ecebbd99f/vendor/plugins/cache_contrib/lib/active_support/cache/patches/auto_load_missing_constants.rb It''s not really a bug per se, just a limitation in that un-marshaling doesn''t autoload. It''d be nice to see Rails.cache support this type of behavior though. On Jan 25, 2009, at 3:44 PM, ara.t.howard wrote:> > > i''m seeing issues with how caching interacts with auto const loading > in production. it''s find if the objects cached are top-level > themselves, but an object graph (model + associations) is cached i''ll > get errors when a production process first comes up (no models loaded) > and hits the cache to get said object graph. > > this fix, which is a uber hack, shows quite clearly the bug - i''m > *really* curious if anyone else has seen this behaviour > > the hack: > > vendor/rails/activesupport/lib/active_support/cache/mem_cache_store.rb > > > 24 def read(key, options = nil) > 25 super > 26 @data.get(key, raw?(options)) > 27 rescue MemCache::MemCacheError => e > 28 logger.error("MemCacheError (#{e}): #{e.message}") > 29 nil > 30 rescue ArgumentError => e > 31 message = e.message > 32 raise unless message =~ %r|undefined class/module| > 33 class_name = message.scan(%r/\w+/).last > 34 begin > 35 ActiveSupport::Dependencies.load_missing_constant > ActiveRecord::Base, class_name > 36 rescue Object > 37 raise e > 38 end > 39 end > > > in case it isn''t obvious this simply shows that rails const loading > mechanism is not triggered from a Marshal.load, which makese sense > actually.... > > thoughts? > > a @ http://codeforpeople.com/ > -- > we can deny everything, except that we have the possibility of being > better. simply reflect on that. > h.h. the 14th dalai lama > > > > > >--~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "Ruby on Rails: Core" group. To post to this group, send email to rubyonrails-core@googlegroups.com To unsubscribe from this group, send email to rubyonrails-core+unsubscribe@googlegroups.com For more options, visit this group at http://groups.google.com/group/rubyonrails-core?hl=en -~----------~----~----~----~------~----~------~--~---